Machine learning and rainfall threshold-based assessment of landslide hazards in Vietnam

For an effective landslide hazard assessment, it is essential to accurately predict the occurrence, timing, and magnitude of landslides. This work presents a detailed analysis of landslide spatiotemporal probability and size distribution for a case study Vietnam. Spatial probability was modeled using Extreme Gradient Boosting (XGB), Random Forest (RF), and Logistic Regression (LR) with 12 predictor variables and a landslide inventory recorded from 2017 to 2024. Temporal probability was estimated using daily rainfall data, applying an event rainfall–duration threshold in combination with a Poisson model. Landslide size probabilities were derived from a probability density function (PDF). Finally, a set of hazard maps was produced for three different time periods and three landslide size classes.The study has been supported by the Polish National Science Centre (project no 2023/49/B/ST10/02879).